Soil characteristics and their effect on fungal diversity and composition
The principal component analysis (PCA) recovered more than 56% of data variability in the first principal component axis (PC1) in both physical and chemical properties. The PC1 of each PCA was used in further analyses (Fig. 3). In our PCA for physical characteristics, the negative values represent fine texture soils (silt and clay), which are predominantly present in seasonally flooded forests – igapós and várzeas (Fig. 3A). The campinas had plots at both extremes of PC1, having the plots in Jaú and Cuieras localities with fine texture and the others plots localized in Caxiuanã with coarse soil textures (Fig. 3A). Terra-firme was more spread across different gradients of the soil texture (Fig. 3A). In the PCA for chemical compounds, positive values in PC1 represent low-fertility soils. Campina and terra-firme were more associated with low-fertility soils, while várzea forests showed different fertility levels (Fig. 3B). Plots in igapó forests also showed low soil fertility except for the plots in Benjamin Constant (Fig. 3B). For details of soil characteristics see Ritter et al11.
Only the mineral soil had some soil properties with significant effect on the OTU Shannon diversity, an effect that varied by marker (Table 1). For 18S, only the organic carbon (C) content was significant, with a negative effect. Organic carbon was also significant and negative for soil ITS diversity. Chemical PC1 was significant for COI and ITS soil diversity, with a higher effective number of OTUs increase following decreasing soil fertility. The pH and soil texture had no significant effect on OTU diversity.
Geographical distance was significant for all datasets. However, since juxtaposed localities are usually similar in many respects, we cannot differentiate the level of spatial correlation from the effect of soil properties in our analysis of community turnover (Table 2). For community turnover, organic carbon and pH were significant for all soil communities (18S, COI and ITS), as was pH for all litter communities. Organic carbon was also significant for the COI litter dataset. Soil texture was significant in all communities except for the ITS soil dataset (Table 2). The PC1 for chemical properties was significant for the 18S and COI litter communities. In the PERMANOVA analysis, the soil properties were all significant with a weak effect on all datasets (Table S3).
The soil layer, organic litter, and mineral soil had a weak but significant effect on the number of OTUs (PERMANOVA results: p < 0.001 for all datasets, 18S – R2 = 0.05, COI – R2 = 0.04, and ITS – R2 = 0.03). There were small differences between the soil and litter communities in the two axes of non-metric multi-dimensional scaling (NMDS) in all datasets (Fig. 4). The litter COI and ITS datasets had a higher mean number of OTUs, where a higher number of OTUs is considered litter indicators (OTUs with a significantly higher probability to be found in litter than soil; Table 3), and a high number of exclusive OTUs than 18S (Fig. 5). For 18S, the results contrast with those of the other markers, showing soil as the most diverse substrate, with the highest number of exclusive and indicator OTUs (Table 3, Fig. 5C). The majority of indicator OTUs for both layers are saprotrophs (Table S2).